Thus, among the points ‘intended merely to prove that the machine must be regulated by mind,’ we find a telling assessment of the capability of machine reason. ‘The Automaton does not invariably win the game,’ Poe reminds us. ‘Were the machine a pure machine this would not be the case—it would always win.’ He regarded that evidence on the premise that ‘the difficulty of making a machine beat all games, is not in the least degree greater, as regards the principle of the operations necessary, than that of making it beat a single game.’
Standing at the other end of videogame history, we may be inclined to laugh at Poe’s assurance, but his premise is remarkable less for its lack of prescience than for the way it leads him into contradiction. Having already established that ‘[n]o one move in chess necessarily follows upon any one other,’ and that ‘[a]ll is then dependent upon the variable judgment of the players,’ it is difficult to account for his supposition that even a reasoning machine could win in all cases. Given the indeterminate nature of play, it ought always be possible for an opponent to lose, even if that opponent is free of the disorienting passions. That, in fact, is the very feature that recommended a game like chess as a higher standard for machine intelligence than Babbage’s Difference Engine. How then are we to understand the apparent mistake in Poe’s reasoning, save as a demonstration of his faith in the perfectibility of reason and his mournful attitude toward the mind?
The Turk’s Descendants The better part of two centuries have not sufficed to demonstrate the full extent of the difficulty involved. The closest we may have gotten to Poe’s perfected reason may be the 1997 victory of IBM’s chess-playing computer Deep Blue over the world champion Garry Kasparov. Yet, even despite its capacity for evaluating 200 million positions per second, the machine lost the first game of the match and played three more to a draw. Adjustments were made to the machine between games, which might be said to count against its status as pure machine. Nor was its final victory without controversy. Kasparov has gone so far as to imply that human agents surreptitiously intervened during the course of play, their ostensible motive being to ensure a victory that would boost investor interest in IBM stock. If true, that would suggest that we have yet to fully tease the man out of the cabinet. In the meantime, the stakes have changed. Whereas Poe mined a deep vein of horror in the capacity of the human passions to undermine the achievements of civilization, the technological extension of human destructiveness during the 20th century has taught us to regard strict logic with greater skepticism. One particularly apropos expression of that concern is HAL9000, the onboard computer in Stanley Kubrick’s 2001: A Space Odyssey. What at first appears to be a glitch in HAL’s programming is ultimately revealed to be a coldly rational program of murder decided upon in order to protect the mission from the ship’s human inhabitants. Before the killing begins, though, HAL is shown besting one of the astronauts at a game of chess. By way of chess-playing computers like Beep Blue, HAL is both a descendent of the Turk and an objection to Poe’s faith in the perfectibility of logic. At the same time, the scope of machine-played games has expanded vastly. The great virtue of Babbage’s Analytic Engine was its use of punch cards to abstract input from
Memory Insufficient volume one issue five